| 摘要 |
<p><P>PROBLEM TO BE SOLVED: To suppress deterioration in voice recognition performance by performing as much as possible, voice recognition using a dictionary with few unnecessary words included in recognition vocabulary. <P>SOLUTION: A speaker is identified based on voice data, and the using frequency of the unnecessary words in the voice uttered by the speaker is calculated and is stored as the unnecessary word using frequency for each speaker. The unnecessary word using frequency shows the tendency as to how frequently the unnecessary words are used for each speaker. Therefore, when the speaker is identified based on the voice data, and, when the unnecessary word using frequency is stored for the speaker, an unnecessary word dictionary 313b used by a voice recognition section 31 is switched to an unnecessary word dictionary 313b where the number of unnecessary words corresponds to the stored unnecessary word using frequency. As a result, the voice is recognized using the unnecessary word dictionary 313b where the number of unnecessary words is appropriate for the unnecessary word using frequency of the speaker. <P>COPYRIGHT: (C)2011,JPO&INPIT</p> |
